Help Game Designers Analyze Data: How to Develop an Online Pivot Table Tool

[Music] well thanks for the introduction and hello everyone welcome to our talk today the topic of our talk is help game designers analyze data how to develop an online pivot table too so let's get started before the talk here is a quick introduction of myself and my partner i am jjie a project manager from netease and i'm also a game designer in naraka bladepoint team and my partner jung yuan is a2 developer in centerfire ux he has developed a virus of tools from web tools to client tools and he has a coding experience for over 10 years and the two he had developed have shipped in several games in our studio

such as revelation online and naraka etc so today we are here to share how to help game designers quickly get started with decision making through data and cultivate data driven culture by developing a reproducible online pipeline table analysis too with a low development cost we will organize our talk into four parts the first would be what problems we are facing with and the second second part would be the design pattern and how to quickly achieve these two the third part would be a simple demo shows how the tool is chipped in our game and how it works and the fourth part will be the takeaway and suggestions we can bring to all of you okay here comes the first part what problems we are facing with in our studio data-driven decision

making has been a routine among designers data is the most objective factor that loyally reflects the exact operation pattern of a player then designers and the user researchers can dig the real demands hidden behind the behavior in order to cut with data thinking among designers and make decisions quickly through data designers often need to raise demands to our data engineers to obtain data and then use excel or other tools after obtaining the data for analysis however data engineers in our studio often have other important tasks to handle they are often assigned in in-depth research and the development of some important topics in games such as financial system monitoring and social system monitoring also they are responsible for in-game recommendation system and abnormal monitoring

here's a diagram shows how data demand works before our tool chipped in our workflow normally there are three ways to handle a data demand in our studio first is developing a real-time real-time check monitor and however it costs with uh it costs much developing resources second we have a business intelligence platform designers can raise the data demand and use the data dashboard for stable demand and third is directly risk demands to a specific engineer for service a data product manager might be helpful to carry clarify designers demands into keys and rules of the database to the engineer the process will cause several colleagues for every demands and the communication cost is also high let's take and let's take this diagram as an example for a data demand a real-time check

method is instant but it's resource costly the second one is bi platform let's assume it may cost three hours for an engineer to develop for a single demand and the third one is manually fetched it may cost three hours and many colleagues to coordinate to finish the at the fetch operation additionally all of these three methods have communication cost we have data product managers to handle the communication process with designers the communication process may also cost all the relevant people for let's just say for an hour so every method have either resource cost issue or time cost issue but our tool as a customizable tool can handle different demands in one two and a single designer can handle this facing with a situation above hour 2 is targeted to finish the whole data demand process in 30 minutes and only with a

designer his own this diagram shows the traditional way we handle a data demand it takes at least three percent to handle a single demand four processes are involved in it and this diagram shows the target of our tool designers directly get the data through our tool skip the communication process and the repeated operation of writing sql code engineers can spend more time on more valuable work before our tours develop their full process in the total to finish the data demand but after we ship it in our tool there are only two processes in total again let me emphasize that developing every tool is for upgrading the efficiency of our colleagues lower both time cost and labor cost and the pilot table 2 is not excluded and then we goes to the second part what

solutions we suggest for an engineer querying data is just writing several lines of sql code to look up some data in the databases but for a designer to accept to expect every designer have a technical background is not practical let them learn single code is not a wise choice the key problem is how to bridge the designers with the data before the tool is developed our data product manager and engineer who communicate with designer and the marketing marketing department had really a hard time to know what some designers really want like one of our data product manager mentioned this demand is a total tragedy and a long word document with ambiguous descriptions finally i found that they even did not

tell me which locks to search uh i do not know which designer gave these demands but it is obvious that the process cost cost much time for designers to clarify with data product manager and engineer about the exact demands they want gradually our data engineers work out a regulation to let designers to follow designers are required to provide an excite file including important messages and rules they want they want to add to the sql code like in this picture they are told to provide time information like every friday saturday sunday from march 11th and the logs they want to check is the gm trials dungeon reward log and the inner logic of some data like like the average daily activity it costs

less time in communication but still need engineers to handle the sql search operation still more than one people are involved in this process following this we want a more direct way to let designers to handle data demand on their own something like advanced search engine it should have powerful functions and customizable keys but advanced search is a tool which is hard for designers to learn too many cases to add and it is easy to miss important information we should not give designers too much things without directions we need fewer entrances but lead to a correct sql code with enough messages so after this with a more practical method is to translate the sql code into natural language that every everyone can understand designers can set the keys like filling

some forms to finish the paragraph of text and a sql language can be generated automatically to automatically query the results in the databases so combining with the data structure contained in our game we decided to use a 5w model to design a question and answer system the designers can use 5w to organize their data demands in several questions so 5w2h is originally a concept in management field to disassemble an event into several elements we modified the date from original ones to meet our situation so r5w is who which refers to the primary key of this data query and whom which refers to the detailed description of the primary key and what refers to the purpose of the

data quality when refers to the range of the data query and where refers to the server range selected for this query we use 5w to cover all keys a designers can query they just need to choose exact keys among several drop down menu and 5w can disassemble demands like disassembling other events and it only provides five entrances for designers to fill it won't be too messy for them to handle and let us see how exactly this method works this sentence shows the designer want to query all rows in november 16 at all servers all channels the information include the rows name date they create the row and the servers and the occupations and the battle power the

5w questions will show one by one for designers to answer it is also a good way to ask designer to think again about their data demands by go back through the sentence again why this mean we try to let the designer know what are the elements to form the data demands and this is how the result of our two given we provide the data through a pivot table designers can modify the pivot table with the exact key they want to show and digital visualization is another important demand of our two designers need an intuitive view of our data a pivot table is a good way to representing data with more than two keys so that is why we decided to aggregate the private table into level two so enough part for the design pattern and now it comes to the engineering part

let's leave the stage for my partner jung yuan he will introduce how to achieve such kind of tool with low cost method however i move on in time how to achieve that have a table to present day is a natural issue with our facebook ground development let's see the test stack of our tool first as a terrible web tool they run an user mainstream technology structure top screen and the layout the component level usually open source level and the design the aviation is the python jungle which we can implement lately and easily and the database in my circle

so how to achieve the private table based on our stack so we image investigated some of the public technology solutions commercial software like flex master dhtmr pirate is not suitable for us since it costs more money open source components such as the power treble js are not suitable for us and because the ui is not clearly usually friendly enough and it's not easy to extend so we decided to use the terrible components of any sun to accent the information after we decided to implement this and design table there are there some collapse issues the backend man missed two problems performance of expression date

and position of circle the front and the main missed performance six year number had an added possession frozen calm blue and the survive i'm here to share three members i have in common to you [Music] the first line is how to implement those serial numbers on the top and the left of the table many options private table components don't have a cell number for the private table but itself has the same number so our designer also wants the number so again with the same tools and relations we also make r2 intention it's very similar to use and then to implement this we all need to see to set the first column as a number of sprints

column and the number size can be increased according to the hidden length to achieve the number of screens for the order of the top we can set the letters to the value of the header if there are more than 26 commas we will use a bb the second is table header setting and the data procession our table are generally used to view straight mission data rows elements and volume adrenal tables can only view two dimensional data so we need to make sure and decision table component via metal dimension there use the fusion that the header of the anagen table components can be nasty i put the value of the column in header and through the national header

and let the low and validate spread normally so that the most basic globe come and the variation of the power table will be quickly displayed in our table many people may think that the headquarters only put the things back titles but we have to read the commission thing so that you can achieve some expected results here is how i extremely said in children of the commons the length the limitless and the results of the final power table are shown in the figure blog the third problem is freeze to laws and the columns of the table

once there is a lot of data in the table the current loss on this column can be floating after the select suddenly which is committed for display the data that's also the reason why i didn't choose other resource power table components because they don't provide this function and it's very chopstick to implement the energetic components the birth rating comes by default we only need to monitor the caliber use and use the sales card already to face this loan the front is the effect f after we add the road links and there's no need to worry about the day we want to check after too much scoring after brief resolution of the design

pattern and a new pattern of our term let's see the exact situation when are two chipping out games let's take a liberation example the ocean line is a memory mobile game with large number of bales promote the buy our studio due to the large scale of playoffs massive has been generated there are designers are accustomed to the joint decision making and it has become a future of the development group they also brought a large part of our daily date month as a household designers use our tools i too have been tuned into the game debris designers now the left users are true for dating months up to now we have children in short games and and make future

contributions in more games by having nana slowly and annoying status easily part four how to improve empire table in your studio today we will show how to have a game designer quick get started with the position making slow date and cultivate a german card by developing a reducible online power table analysis to reach a low development coaster

we also showed what is variable to us leader and how it changed the workflow and if you want to implement a two lakh or about table i can give you two suggestions here first try to cut the dead german source among your designers own your strong way of employing user experience through that and make decisions so that can to make better confusions secondly the purpose of developing truth is to improve if reduced repetitive men level and reduce causes the directing purpose of r2 is the same we must remove the using of sources of law for lecture development the technologies used by autos are here different if you are interested in gym you can

also live to implant making a simulator in your short cloud of time and here is segment rate that you may be interested in first the electro is available in design and dates second finder can have designers to replace their advance and cultivate their journey thinking so i provide a secret and the low custom tool that developers can refer especially for innovative developers thank you for attendance [Music]

you